Plant-inspired behavior-based controller to enable reaching in redundant continuum robot arms

TL;DR

The paper tackles enabling reaching for highly redundant continuum arms in unstructured environments by replacing vision-heavy task-space controllers with a plant-inspired, behavior-based framework that relies on embedded proximity sensing. It introduces a bottom-up control scheme combining primitive tendon-based bending with abstract behaviors (circular shifts and learning from history) to emulate growth-driven movements, including a circumnutation-like exploration phase followed by a proximity-guided reaching phase. The approach is validated on a 9-DoF modular, cable-driven arm, demonstrating characterization of bending directions, exploration to locate targets within the observable space, and improved reaching performance when antagonist tendon pulling is proportional to curvature. The findings suggest that embedded sensing and distributed control can extend the deployability of continuum and soft arms to unstructured settings, with practical implications for robotic manipulation where vision is limited or unreliable.

Abstract

Enabling reaching capabilities in highly redundant continuum robot arms is an active area of research. Existing solutions comprise of task-space controllers, whose proper functioning is still limited to laboratory environments. In contrast, this work proposes a novel plant-inspired behaviour-based controller that exploits information obtained from proximity sensing embedded near the end-effector to move towards a desired spatial target. The controller is tested on a 9-DoF modular cable-driven continuum arm for reaching multiple setpoints in space. The results are promising for the deployability of these systems into unstructured environments.

Figures (9)

• Figure 1: The (A) modular continuum robot arm comprises of a concatenation of (B) activation units. The overall system is a combination of a pair of activation units that are independently activated, and is mounted on a (C) custom-designed rigid base. The system monitors the environment through a distributed arrangement of (D) proximity sensors placed in distal portion of the arm.
• Figure 2: The antagonist restoring length is proportional to the agonist pulling length.
• Figure 3: Overview of the behaviour-based control architecture.
• Figure 4: The principal bending directions are identified by pulling respective tendons. (A) Comparison between bending angle and structure curvature over all the bending directions. (B) Segment trajectories while moving along the bending directions.
• Figure 5: Targets are placed around the continuum arm to show the controller invariancy to their position. The closest reaching continuum arm configuration is reported for each target. For sake of visibility, targets are represented with half of their actual diameter.